Fusion energy aims to generate electricity by forcing light atomic nuclei together under extreme heat and pressure, releasing energy through the same process that powers the sun. After decades of incremental progress, the field crossed a symbolic threshold in late 2022 when the National Ignition Facility in the United States achieved target energy gain greater than one, meaning the fusion reactions produced more energy than the lasers delivered to the fuel capsule. That milestone energized both public funding and a growing wave of private investment, but turning a laboratory demonstration into a power plant that feeds electricity into the grid remains an enormous engineering challenge, touching everything from plasma physics and exotic materials to tritium fuel supply and regulatory frameworks.
How Fusion Releases Energy
Fusion works by merging light nuclei into heavier ones. The most accessible reaction combines two hydrogen isotopes: deuterium and tritium. When these fuse, they produce a helium nucleus and a high-energy neutron, releasing about 17.6 million electron-volts of energy per reaction. That is roughly ten million times more energy per unit mass than burning coal. The catch is that both nuclei carry a positive charge and repel each other violently. Overcoming that electrical repulsion requires heating the fuel to temperatures above 100 million degrees Celsius, far hotter than the core of the sun. At those temperatures the fuel exists as plasma, a soup of free-roaming ions and electrons that must be held together long enough and densely enough for a useful number of fusion reactions to occur.
The conditions needed for net energy production are captured by what physicists call the Lawson criterion, which sets a minimum combination of plasma density, temperature, and confinement time. If the plasma is too thin, too cool, or disperses too quickly, more energy goes in than comes out. Every major fusion experiment in the world is, at its core, an attempt to beat this threshold by a wide enough margin to make a power plant practical.
Magnetic Confinement and Its Evolving Designs
The most mature approach to fusion uses powerful magnets to trap the plasma inside a doughnut-shaped vacuum chamber. The leading design, the tokamak, threads the plasma with spiraling magnetic field lines created by a combination of external coils and an electric current driven through the plasma itself. ITER, the international tokamak under construction in southern France, is designed to produce about 500 megawatts of fusion power from 50 megawatts of heating input, which would be the first demonstration of a large net energy gain from magnetic confinement. But tokamaks are not the only magnetic option.
Stellarators take a different philosophy. Instead of relying on a current flowing through the plasma, they create the confining magnetic field entirely with external coils twisted into elaborate shapes. This removes the risk of sudden plasma current collapses (called disruptions) that plague tokamaks, but stellarators are harder to design and build. Recent computational work has shown that a property called quasi-symmetry, where the magnetic field strength has a hidden symmetry even though the physical shape of the device does not, can be achieved to extraordinarily high precision. One study demonstrated that for a coil set producing a 1-tesla field, departures from perfect quasi-symmetry could be driven down to the scale of Earth’s background magnetic field, roughly 50 microtesla.1PubMed. Magnetic Fields with Precise Quasisymmetry for Plasma Confinement Follow-up work showed that coils built to these optimized specifications would lose only about 0.04% of energetic particles over 200 milliseconds when scaled to reactor size, a dramatic improvement over earlier stellarator coil configurations.2PubMed Central. Precise stellarator quasi-symmetry can be achieved with electromagnetic coils
These results matter because poor particle confinement has historically been the stellarator’s Achilles’ heel. If the new coil designs hold up in hardware, stellarators could become serious competitors to tokamaks for commercial reactors, trading the tokamak’s disruption risk for more complicated magnets but steadier operation.
Inertial Confinement and the NIF Milestone
Where magnetic confinement holds plasma in place for seconds at a time, inertial confinement fusion (ICF) takes the opposite approach: compress a tiny pellet of fuel so fast and so hard that the nuclei fuse before the plasma has time to fly apart. At the National Ignition Facility (NIF) in California, 192 laser beams deliver a pulse of ultraviolet light to a small gold cylinder called a hohlraum, which converts the light into X-rays that crush a peppercorn-sized capsule of deuterium-tritium ice.
In December 2022 the NIF achieved target gain greater than one for the first time, producing more fusion energy than the laser energy that entered the hohlraum.3Plasma Physics and Controlled Fusion. Present understanding of ignition and gain using indirect-drive inertial confinement fusion target designs on the U.S. National Ignition Facility Subsequent experiments refined the design further. By reducing the fuel adiabat, which essentially means compressing the fuel more gently before the final shock crushes it, researchers achieved up to 80% higher fusion yield and 14% greater fuel compression compared to earlier shots, and the improved design became the only configuration to exceed target gain with less than about 2 megajoules of laser energy.4Physics of Plasmas. First demonstration of improved yield with reduced adiabat in inertial confinement fusion implosions on the National Ignition Facility
A caveat that often gets lost in headlines: “target gain greater than one” means the fusion output exceeded the laser energy hitting the target, not the total energy drawn from the wall socket. The NIF’s lasers are roughly 1% efficient, so the facility as a whole consumed far more energy than the fusion reactions released. Moving from scientific breakeven to engineering breakeven, and then to a commercially viable power plant, requires improvements in laser efficiency, repetition rate (the NIF fires about once a day; a power plant would need several shots per second), and target manufacturing at scale.
The Tritium Problem
Deuterium is cheap and abundant, easily extracted from seawater. Tritium is neither. It is radioactive with a half-life of about 12 years and does not exist in useful quantities in nature. The global inventory comes almost entirely from heavy-water fission reactors in Canada and South Korea, and the supply amounts to only a few kilograms per year. A commercial fusion plant burning deuterium-tritium fuel would consume on the order of 100 kilograms of tritium per year.
The standard plan is for each reactor to breed its own tritium. Neutrons escaping the plasma would slam into a “blanket” of lithium surrounding the reaction chamber, splitting lithium atoms and producing new tritium. In theory, the neutron math works out: each fusion reaction releases one neutron, and adding a neutron multiplier like beryllium or lead can push the tritium breeding ratio above one, meaning the reactor makes slightly more tritium than it burns. In practice, the margins are uncomfortably thin. Every penetration in the blanket for heating systems, diagnostics, plasma exhaust, and structural supports reduces the effective breeding coverage. Some analyses of proposed blanket designs using lithium silicate or lithium titanate breeders with beryllium multipliers have found that tritium self-sufficiency may not be achievable once realistic engineering integration is accounted for.5Fusion Engineering and Design. First principles review of options for tritium breeder and neutron multiplier materials for breeding blankets in fusion reactors
This is one of the least glamorous but most consequential problems in fusion engineering. If a reactor cannot reliably breed enough tritium to replace what it burns plus a small surplus to start up the next reactor, the fuel supply becomes a bottleneck for the entire fleet.
Materials Under Siege
The inside of a fusion reactor is one of the harshest environments engineers have ever tried to build for. The plasma-facing wall absorbs intense heat loads, a constant rain of 14-MeV neutrons, and periodic bursts of plasma during instabilities. Tungsten is the leading candidate for the surface layer because it has the highest melting point of any metal (about 3,400 °C), low erosion rates under plasma bombardment, and does not become highly activated by neutron irradiation compared to heavier elements. But tungsten has a serious weakness: it transitions from ductile to brittle at temperatures between roughly 150 °C and 400 °C, which means that at the water-cooled operating temperatures planned for many reactor designs, it is inherently fragile.6Journal of Nuclear Materials. Tungsten as material for plasma-facing components in fusion devices
Behind the tungsten armor sits structural steel, and neutron damage accumulates there too. High-energy fusion neutrons displace atoms from their crystal lattice positions, creating defects that degrade the material over time. Modeling of a proposed European demonstration reactor (DEMO) estimated that iron and tungsten at the first-wall location would accumulate roughly 1.2 and 0.9 displacements per atom per full-power year, respectively.7Fusion Engineering and Design. Displacement damage study in tungsten and iron for fusion neutron irradiation On top of the lattice damage, neutrons initiate nuclear reactions inside the structural materials that transmute elements and generate trapped gases, especially helium. Helium atoms collect at grain boundaries and cause additional swelling and embrittlement.8Journal of Nuclear Materials. Neutron-induced dpa, transmutations, gas production, and helium embrittlement of fusion materials
No material testing facility today can replicate the exact neutron spectrum and flux of a deuterium-tritium fusion plant. Fission reactors and spallation sources provide approximations, but validating structural materials under true fusion conditions will likely require operating pilot plants themselves, a bit of a chicken-and-egg problem.
How Fusion Compares to Fission on Safety and Waste
Fusion often gets described as “clean nuclear energy,” and on the two biggest concerns people have about fission power, that framing holds up reasonably well. A fusion reactor cannot melt down in the way a fission reactor can. The plasma contains only a few grams of fuel at any moment, and if confinement is lost the plasma cools and the reaction stops within milliseconds. There is no chain reaction to run away. Fusion waste contains no transuranium elements (plutonium and its relatives) and no fission products like cesium-137 or iodine-131.9Fusion Engineering and Design. Waste from fusion reactor: A comparison with other energy producing systems
What fusion does produce is neutron-activated structural material and residual tritium. The activated steel and tungsten are radioactive, and in the first years after shutdown the total radiotoxicity of a fusion reactor’s waste can actually exceed that of a fission pressurized water reactor because of short-lived activation products. The difference is how fast the hazard fades. After about a century of decay, the radiotoxicity of fusion waste drops to more than 100 times lower than that of fission waste from a conventional reactor.10Fusion Engineering and Design. Fusion power plants, fission and conventional power plants. Radioactivity, radiotoxicity, radioactive waste Most fusion waste is expected to qualify for shallow land burial rather than deep geological disposal, though that classification depends on careful material selection during reactor design.
There is one genuine plasma-physics hazard worth understanding: runaway electrons. During a sudden loss of plasma confinement (a disruption), the rapidly changing magnetic fields can accelerate a population of electrons to near light speed. In a large tokamak, these relativistic electron beams can carry tens of megaamps and, if they strike the wall, cause severe localized damage. Modeling of a proposed spherical tokamak reactor called STEP found that unmitigated disruptions could convert the entire plasma current into runaway electron beams of 10 megaamps or more.11Nuclear Fusion. Disruption runaway electron generation and mitigation in the Spherical Tokamak for Energy Production (STEP) Engineers are developing passive safety systems, such as in-vessel helical coils that automatically generate magnetic perturbations during a disruption, scattering the runaway electrons before they can concentrate into a damaging beam.12Nuclear Fusion. Passive deconfinement of runaway electrons using an in-vessel helical coil The appeal of passive systems is that they need no external power or disruption detection; the disruption itself drives the coil current.
Teaching Machines to Predict Disruptions
Even with passive mitigation hardware, the preferred strategy is to see disruptions coming and prevent or soften them. This is where deep learning has made striking progress. A 2019 study trained a neural network on experimental data from two major tokamaks, DIII-D in the United States and JET in Europe, and demonstrated that the trained model could predict disruptions on a machine it had never seen before.13Nature. Predicting disruptive instabilities in controlled fusion plasmas through deep learning That cross-machine capability matters enormously for ITER, which will not be able to afford thousands of disruptive shots to train a model from scratch.
More recently, a transfer-learning approach trained on a small Chinese tokamak (J-TEXT) was successfully transferred to the much larger EAST tokamak using only about 20 sample discharges, achieving comparable prediction accuracy to a model trained directly on EAST with roughly 1,900 discharges.14Communications Physics. Disruption prediction for future tokamaks using parameter-based transfer learning Separately, a deep-learning predictor has been deployed in the live plasma control system of the HL-2A tokamak, where it achieved about 89% total accuracy and was used to trigger emergency gas injection. Over 80% of disruptions were predicted with enough lead time (more than 12 milliseconds) for the mitigation system to act.15Fusion Engineering and Design. Real-time disruption prediction in the plasma control system of HL-2A based on deep learning
Twelve milliseconds sounds vanishingly short, but in a hot plasma, that is an eternity compared to the microsecond timescales of the disruption’s thermal collapse. The trend is toward faster, more generalizable models that can be layered into reactor control systems alongside conventional feedback loops for position, shape, and heating.
Alternative Fuels and Hybrid Concepts
Deuterium-tritium fuel is the lowest-hanging fruit because it has the highest reaction rate at achievable temperatures. But it comes with the tritium supply problem and produces copious high-energy neutrons that damage everything they hit. Aneutronic fuels, which release most of their energy as charged particles rather than neutrons, have obvious appeal. The most discussed candidate is proton-boron-11, which produces three helium nuclei and no neutrons in the primary reaction.
The physics, however, is punishing. A recent evaluation of the Lawson criterion for proton-boron-11 found that net energy production is only possible when the electrons in the plasma are kept significantly cooler than the ions, because hot electrons radiate energy away as bremsstrahlung faster than the fusion reactions can replace it. Even under favorable assumptions about the electron-to-ion temperature ratio, the minimum confinement conditions required are roughly an order of magnitude more demanding than for deuterium-tritium.16Frontiers in Nuclear Engineering. Evaluation of the Lawson criterion for aneutronic proton-boron-11 fusion: effects of ion temperature and bremsstrahlung losses Proton-boron-11 fusion is not impossible, but it is a much harder target, and no experiment has come close to the required plasma conditions.
Between mainstream magnetic or inertial confinement and the far frontier of aneutronic fuels sit hybrid concepts. Magnetized target fusion, for instance, forms a compact plasma with embedded magnetic fields and then mechanically compresses it to fusion conditions. Field-reversed configurations, a type of compact plasma, have been explored for this purpose because they can be formed, translated, and compressed relatively simply compared to a full tokamak.17Review of Scientific Instruments. FRX-L: A field-reversed configuration plasma injector for magnetized target fusion Several private companies are pursuing variations on this theme, betting that smaller, simpler machines can reach fusion conditions without the multi-billion-dollar scale of ITER.
Economics and Regulation
Even if the physics and engineering problems are solved, fusion must compete on cost with renewables, fission, and natural gas. An economic model of a hypothetical European demonstration reactor estimated a levelized cost of electricity around $160 per megawatt-hour, with a break-even selling price near $175 per megawatt-hour.18Energy. Approximation of the economy of fusion energy For context, onshore wind and utility-scale solar in many markets already come in below $50 per megawatt-hour. Fusion’s value proposition rests less on being the cheapest electron and more on providing firm, dispatchable, low-carbon baseload power that does not depend on weather, geography, or geological storage of carbon dioxide.
On the regulatory side, a key question is whether fusion plants will be regulated like fission plants, with the heavy (and expensive) licensing burden that entails. The United Kingdom has taken the most explicit position so far, deciding that future fusion facilities will be regulated under existing health and safety and environmental law rather than under the nuclear-specific regulatory regime used for fission reactors. The decision was based on the judgment that the hazards involved in fusion are proportionally lower than those of fission.19Nuclear Fusion. Recommendations for the future regulation of fusion power plants The United States has moved in a similar direction, with the Nuclear Regulatory Commission signaling that fusion devices will not require the same class of license as fission plants. How other countries handle this will significantly affect the cost, speed, and commercial attractiveness of building fusion power stations.
Hydrogen Production and Other Non-Electric Uses
Electricity generation gets most of the attention, but a commercial fusion reactor would also be a potent source of high-temperature heat, which opens the door to industrial applications that conventional nuclear and renewable plants handle poorly. One of the most studied is large-scale hydrogen production. Thermochemical water-splitting cycles, such as the sulfur-iodine process, need sustained temperatures above 800 °C and would pair naturally with the high-grade thermal output of a fusion blanket.20International Journal of Hydrogen Energy. Hydrogen production using fusion energy and thermochemical cycles
A modeling study of a helical fusion reactor operating with molten-salt mixtures estimated hydrogen production rates as high as 40 kilograms per second when coupled to thermochemical or high-temperature electrolysis processes, alongside fissile fuel breeding as a side benefit.21International Journal of Hydrogen Energy. Hydrogen production via water splitting process in a molten-salt fusion breeder Those numbers are from a specific theoretical design and should not be read as guaranteed output, but they illustrate the scale of what fusion’s thermal energy could enable. In a future economy hungry for green hydrogen to decarbonize steelmaking, ammonia production, and long-haul transport, a dispatchable high-temperature heat source would be genuinely valuable even if the electricity it generated were not the cheapest on the grid.
Other potential non-electric applications include desalination, district heating, and the production of medical isotopes. Whether any of these become practical depends on whether the first generation of fusion plants can demonstrate reliable operation and begin driving down costs through iteration, a cycle that fission power never fully completed because of public opposition and regulatory stagnation. Fusion’s window of opportunity may depend as much on political and economic timing as on the physics.